A report on the status of phishing on the web... sample of technology initiatives being used to stem

A report on the status of phishing on the web... sample of technology initiatives being used to stem

A report on the status of phishing on the web and
sample of technology initiatives being used to stem
the tide of phishing attacks
UNISA-TR-2006-03
Authors: Upasna Bechan
Supervisor: Alta van der Merwe
1
Contents
CONTENTS
2
LIST OF FIGURES
3
GRAPHICAL REPRESENTATION OF THE REPORT
4
INTRODUCTION
5
WHAT IS PHISHING
7
WHERE DID IT ORIGINATE
7
WHAT FORMS OF PHISHING ARE THERE?
8
HOW DOES IT WORK?
11
HOW TO MODEL PHISHING ATTACKS?
13
WHY DO PEOPLE FALL FOR IT?
15
THE ANTI PHISHING MEASURES IMPLEMENTED BY YAHOO, EBAY AND
EARTHLINK
17
SCANNING AND ALERTING SOFTWARE TO DETECT AND DEFEND
AGAINST PHISHING ATTACKS
20
EMAIL VERIFICATION TOOLS
23
WEBSITE VERIFICATION SOLUTIONS
25
GUARANTEE OF LEGITIMACY OF MESSAGES AND THEIR SOURCE
28
TWO FACTOR AUTHENTICATION FOR CUSTOMER ACCESS TO
FINANCIAL SERVICES
30
IDENTITY THEFT
34
CONCLUSION
37
REFERENCES
39
2
List of Figures
FIGURE 1, GRAPHICAL REPRESENTATION OF THE REPORT LAYOUT ........................... 4
FIGURE 2. NEW PHISHING SITES BY MONTH OCTOBER-AUGUST .................................. 5
FIGURE 3. PASSWORD STEALING MALICIOUS CODE URLS ............................................. 6
FIGURE 4. PAYPAL PHISHING EMAIL.................................................................................... 9
FIGURE 5. THE PHISHING SITE LOADS UP AN ADDRESS BAR SPOOF AND A MIRROR
OF THE LEGITIMATE PAYPAL.COM SECURE LOGIN PAGE .................................... 10
FIGURE 6. DETAILED PHISHING ATTACK........................................................................... 13
FIGURE 7. MODEL OF A PHISHING ATTACK ...................................................................... 14
FIGURE 8. EXAMPLE OF A PHISHING ATTACK.................................................................. 15
FIGURE 9. PHISHING TOOLBAR: EBAY ACCOUNT GUARD.............................................. 17
FIGURE 10. THE EARTHLINK SCAMBLOCKER TOOLBAR................................................. 18
FIGURE 11. EARTHLINK TOOLBAR SITE WARNING FEATURE ........................................ 19
FIGURE 12. SPOOFGUARD ARCHITECTURE ..................................................................... 22
FIGURE 13. SPOOFGUARD TOOLBAR ................................................................................ 23
FIGURE 14. STRONG TOKEN-BASED AUTHENTICATION ................................................. 31
FIGURE 15. MAN-IN-THE-MIDDLE ATTACK STRUCTURE ................................................. 33
FIGURE 16. . IDENTITY THEFT LIFECYCLE ........................................................................ 35
3
Graphical representation of the report
This report comprises 8 sections with Section 8 being a utility section for
references. Sections 5 and 6 can be read in any order as their content is
confined to their section boundaries.
Figure 1, Graphical representation of the report layout
4
Introduction
This report begins with a description of what phishing is, how a phishing
attack is conducted and why users fall for these attacks. It then goes on to
explore a selection of technology initiatives being investigated in an attempt to
handle the rising number of phishing attacks. The six areas of initiatives
explored are:ƒ
The anti phishing measures implemented by Yahoo, eBay and
EarthLink
ƒ
Scanning and alerting software to detect and defend against phishing
attacks
ƒ
Email verification tools
ƒ
Website verification solutions
ƒ
Guarantee of legitimacy of messages and their source
ƒ
Two factor authentication for customer access to financial services
According to the Phishjing Activity Trends Report (August 2005), the number
of unique phishing websites detected by the Anit-Phishing Working Group
(APWG) was 5 259 in August 2005, the highest number this far. This statistic
helps to highlight the fact that phishing is a lucrative crime that will continue to
grow unless effective counter measures are sought and implemented.
Figure 2. New Phishing Sites by Month October-August
5
It is disturbing to note in the Phishing Activity Trends Report (November 2005)
that the theft of passwords is increasing. This observation leads one to fear
that the theft of passwords will lead to the theft of personal information that is
accessible once the password is known.
Figure 3. Password Stealing Malicious Code URLs (APWG 2006, p. 6)
The illegal access and use of personal information is referred to as identity
theft. Identity theft is discussed in Section 6 according to the following broad
guidelines:•
what kinds of identity theft there are
•
what can it be used for.
The remainder of this paper is organized as follows. Section 2 begins with
background information on phishing. Section 3 looks at the mechanics of how
a phishing attacks is conducted. Section 4 attempts to understand why people
are lured into a phishing attack. Section 5 explores a number of
countermeasures and their effectiveness against the attack they are being
used against. Section 6 is a brief introduction to problem of identity verification
and includes a theoretical architecture that has been proposed in response to
the identity theft problem. Section 7 is the conclusion and Section 8 contains a
list of references.
6
What is phishing
Phishing is the term used by the online community to describe a form of social
engineering whereby unsuspecting users divulge personal information (such
as credit card numbers, usernames, passwords and social security numbers)
to 3rd parties. The 3rd party then uses this information to commit credit card
fraud and identity theft leading to financial losses by the owner of said
personal information.
According to Rudd (2004), the term ‘phishing’ originates from the word
‘fishing’ where the fishermen use bait to attract fish like the criminals use
email to attract online victims. The ‘f’ was replaced by ‘ph’ in recognition of
‘phreaking’, an early hacking method where the hacker would use someone
else’s phone line for their own use.
Where did it originate
The earliest cited phishing attack is the one launched against AOL in the early
nineties where criminals stole AOL passwords. Morrison and Nuttall (2004 :1)
say “Phishing first came to light in 1996 when AOL subscribers began
receiving bogus instant messages asking for their log-on passwords in order
to update files. The aim was to avoid paying a subscription by using someone
else’s account …”
Phishing was seen as largely a novice pastime but recently the activity seems
to be more organised and launched against organisations for major financial
benefit. Kerner (2004) quotes the Gartner survey of 5000 individuals which
estimates the cost of damage to US banks and credit card issuers at 1.2
billion US dollars. There could be further financial implications as consumers
lose confidence in transacting online.
7
What forms of phishing are there?
Identify theft can occur by a variety of techniques. The most widely known and
most visible technique is sending the user an email which contains a link to a
fraudulent website. Once the user is convinced to click on the link (See the
section ‘Why do people fall for it?’ for a discussion on how this is
accomplished), they are redirected to the spoofed site where they enter their
personal details. The details are then sent to the criminal’s web
server/database.
Below is an example of an email from APWG (2005) that was used in a
PayPal phishing attack.
8
Figure 4. PayPal phishing email
The link referred to in the email above takes the user to the spoofed website
listed below:-
9
Figure 5. The phishing site loads up an address bar spoof and a mirror of the legitimate paypal.com secure login page
10
A variation on the email scam described above is to embed an html form in
the email that is sent out. The user then enters their credentials in the email
form which has been embellished with logos/colour schemes to accomplish
the appropriate branding.
Another effective technique to illegally obtain user information is to send the
user a worm that installs key logger software on the victim’s personal
computer (PC). All input data is then sent to the criminal’s website where the
required information is filtered out. This technique is more difficult to detect
hence the user is unaware that anything is wrong until fraudulent credit card
transactions, etc. occur.
A further technique is to create a pop up window over the legitimate website.
The user then enters their login credentials on the pop up thereby giving the
criminals the information they need. The user does not realise what has
happened as they still have the legitimate website in front of them.
A recent trend in phishing attacks is targeting Internet Relay Chat systems
(IRC’s). During a chat session the user downloads a file with attachments that
install key logging spy ware on his personal computer (PC).
How does it work?
Below is a detailed analysis of what happens during a phishing attack from
Tumbleweed Whitepaper (2004 :3,4):Assume the following actors and components of the attack:
• The phisher (P)
• The recipient (R)
• The company whose domain is being spoofed by the phisher (C)
• The phisher’s email server (Ps)
• The recipient’s email server (Rs)
• The recipient’s email client (Rc)
11
1. P generates a fraudulent email with content that looks just like a legitimate
email from C to all its customers. The colors, graphics, text treatment, and
composition are identical to what C uses to normally contact its customers.
There is no way for C to technically prevent P from creating this type of
content. The message in this email is particularly insidious as it describes
how “A recent set of phishing attacks have corrupted our customer account
database at C. Please help us reinstate your account at a secure website
provided by the C Security Service by clicking on the link below”. P then
inserts the email address customerservice@C.com in the FROM field of
the email.
2. P sends this email to as many email addresses as he can get a hold of
using server Ps. He may have done previous spam attacks to understand
which email addresses are likely C’s customers to reduce the number of
phishes he must send to be effective. The domain of Ps happens to be
Csecurityservice.com. The address P uses in the MAIL FROM address of
the email is customerservice@Csecurityservice.com. This is so that any
bounces from invalid recipient email addresses are sent to P’s servers and
not C’s. This will prevent C from being Joe Jobbed and noticing a large
number of bounces coming from email it knows it did not send.
3. R happens to be a customer of C and has heard about these phishing
attacks on the Internet. R’s email server (Rs) receives a connection from
Ps, accepts it, and stores the message in R’s email inbox. R uses his email
client (Rc) to download the message. When he sees P’s email, he inspects
it closely. The address that Rc displays, customerservice@C.com, has
C.com in it, and the “customerservice” string to the left of the “@” matches
another email he got from C a few months back regarding a product return
he made. R trusts this address as legitimately coming from C.
4. While R thinks he’s savvy to phishing attacks, the content of P’s email
takes him off-guard. “Why sure”, R thinks to himself, “These phishing
attacks probably are wreaking havoc on corporate account databases. I
read about it in the papers. I better make sure my account is not ruined,
because I use C’s site often.” R clicks on the link and is taken to a Web
site.
5. R is taken to https://www.Csecurityservice.com/accountreinstatement. R
remembers that the email said the site to reinstate his account was
provided by a “C Security Service”, so the syntax of this URL makes
perfect sense to him. The SSL connection his browser tells him has been
made further convinces him that the site is legitimate. The form on the site
asks for just the right kind of personal information that C would ask of its
customers in order to reinstate an account. R fills out the form and hits the
submit button. The phisher has now succeeded in stealing personal
information from R.
6. R realizes some time later that he’s been phished. C may discover P’s
attack from R or other victims’ reports. By the time C is able to get P’s ISP
to shut down the https://www.Csecurityservice.com/accountreinstatement
12
URL, P has already gathered hundreds if not thousands of sets of
passwords and other personal account information. The fact that P can
never again use https://www.Csecurityservice.com/accountreinstatement
as a Web site or even customerservice@Csecurityservice.com as the
MAIL FROM address in another phishing attack (due to updated blacklists
in anti-spam servers) is beside the point. He’s made a killing with stolen
identities he already has. P can easily move on to the next company’s set
of customers to launch a phishing attack. He might even try phishing C’s
customers again with an entirely different set of legitimate-looking domain
names and email content.
Figure 6. Detailed phishing attack
How to model phishing attacks?
The purpose of using a model as opposed to descriptions (See the section
How does it work?) is the model is able to capture “a variety of attacks in a
uniform and compact manner”, Jakobsson (2005 : 4).
The graphical model allows the analyst to visually quantify the threat to the
system under review. The phishing attack is modelled by a phishing graph “…
in which nodes correspond to knowledge or access rights, and (directed)
edges correspond to means of obtaining information or access rights from
already possessed information or access rights … ”, Jakobsson (2005 : 2).
Below is an annotated example of a phishing graph with detailed explanations
of the labels below.
13
conjunction
starting state
target
disjunction
Edge carries description of cost, effort, probability.
Figure 7. Model of a phishing attack
Vertices can correspond to access to either information or a resource; the
distinction in the graph is implicit by the associated description
Actions are represented by edges, so two vertices are connected by an edge
if there is an action that allows a user with access to one vertex to obtain
access to the linked vertex. Two vertices may be connected by multiple
edges, corresponding to different actions allowing the transition between
them; this is referred to as a disjunction in Figure 4 above.
A conjunction of actions occurs when a number of edges can be merged into
one and ends up in one vertex.
Edges are labelled with descriptions of the effort, probability and
circumstances under which the transition will succeed.
The target refers to the vertex the attacker is attempting to gain access to.
Figure 5 is an example of a phishing attack described in the graphical
notation of Jakobsson. The context of the example is as follows:- Access to a
newly opened bank account can be obtained by submitting the date and
amount of the last deposit as opposed to a password. The date, amount and
account number is assumed to be known for this example.
14
married?
account number
payment
refund alone
salary
bank
account
Guess withholding, 401(k).
Figure 8. Example of a phishing attack
Access to the account is represented by vertex v1. Knowledge of the victim’s
salary is represented by vertex v2, and the edge e251 corresponds to
guessing the level of withholding and percentage of 401(k) contributions.
Knowledge of the victim’s marital status corresponds to vertex v3, and the
edge e351 is the probability of the tax refund check being deposited alone.
Vertex v4 corresponds to access to performing a payment to the victim, and
the edge e451 corresponds to the action of performing the payment. Vertex v5
corresponds to knowledge of the account number.
Why do people fall for it?
According to Tumbleweed (2004:1) “… the phisher is counting on the fear,
guilt or general willingness of unsuspecting victims to trust the e-mail’s
contents and follow its instructions”.
Ironically it’s usually the fear of phishing attacks and fraudulent activity that is
cited as the reason for needing the user information to be validated. The user
is also told that it’s a matter of urgency that the update is carried out. This
does not give the user time to validate the email or its contents and the
15
criminal gets the information they need before the website is detected and
shut down.
According to Drake, et. al (2004), users trust the email they receive for the
following reasons:a) the company being mimicked is a reputable company e.g. Paypal
b) the spoofers emulate the company’s visible branding
c) the email or spoofed website may contain links to sections of the
legitimate website
d) the from address appears to be from the legitimate website
People also trust that the certificates authority (CA) has carried out the due
diligence. This may not always be true. According to Hall (2005), there are 3
areas of weakness in this process:a) Manual vetting process would not be able to spot fake documents
hence making it easy for fraudulent companies to obtain certificates.
b) The vetting process was not standardised across states and countries
hence making it more error prone.
c) Some CA’s outsource the identity verification process to dubious
subcontractors who vouch for the business without doing the proper
checks
So when the user looks at a web site and sees the padlock on the toolbar at
the bottom of the browser, this is no guarantee of the business’s validity. What
the user has to do is double click the padlock icon, navigate to the ‘Subject’
section and verify the data fields displayed from the digital certificate. This is
too complicated and more often than not the data therein is not accurate.
The solution to the error prone manual vetting process is Second Generation
Automated Vetting. The reasons cited for this process working are:a) domain (web address) is confirmed by the applicant in real time
b) real time email validation
c) real time telephone validation
d) sophisticated fraud-detection algorithms
16
e) browser display of the web sites confirmed internet address
The anti phishing measures implemented by Yahoo,
eBay and EarthLink
Yahoo proposed the DomainKeys infrastructure that verifies the domain of an
email sender. This verification helps in the war against phishing by helping to
curb the proliferation of spam. Shor (2005) describes DomainKeys Identified
Mail (DKIM) by “Each message gets a signature, which can then be checked
for authenticity by verifying the originating domain and whether the message
has been tampered with.” The process of using DKIM is described by
Germain (2005) as follows. The ISP or e-mail gateway authenticates the
message sender. Then the message must pass a reputation score. Thereafter
the Domain Name Server (DNS) is used to verify that the encrypted email
address signature came from the stated sender.
Shor (2005) describes DKIM as “…help to mitigate address spoofing…but
does not eliminate either (phishing or spamming) entirely”.
The eBay toolbar is a browser plugin whose primary role is to keep track of
auction sites. However the toolbar also helps prevent fraud in the following
manner according to Dhamija and Tygar. “AccountGuard”, a feature of the
eBay toolbar monitors the domain names that the user visits and provides a
warning by changing the colour of the tab on the toolbar. The toolbar, usually
grey, turns green if the user is on an eBay or PayPal site or turns red if on a
spoofed website. This toolbars effectiveness is only as strong as the list of
spoofed websites known to eBay and PayPal.
Figure 9. Phishing Toolbar: eBay Account Guard
17
EarthLink has made available a free anti-phishing application called
ScamBlocker. ScamBlocker is a feature of the EarthLink toolbar. The purpose
of ScamBlocker is to prevent users from accessing phishing sites. According
to Tynan (2004) ScamBlocker attempts to stem the phishing tide in the
following manner. ScamBLocker automatically downloads a list of known
phisher sites, sourced from its own list, EBay (online auctioneer) and
Brightmail (antispam vendor). When the user tries to access a fraudulent site,
ScamBlocker redirects them to a page supplied by EarthLink. The EarthLink
page then gives the user the choice of either proceeding to the scam site or
reporting the incident to the EarthLink abuse team. The abuse team will then
attempt to get the site’s host to shut down the site.
The benefit of using ScamBlocker is the user is alerted to the fact that they
are about to visit a known phishing site. Unfortunately ScamBlocker is not
foolproof as the feature is only as effective as the list it has access to.
The following list of figures from Lininger and Dean (2005) illustrate now the
EarthLink ScamBlocker toolbar works:-
Figure 10. The EarthLink ScamBlocker toolbar
18
Figure 11. EarthLink toolbar site warning feature
Digital Impact (2005) defines an effective anti-phishing solution as one that
meets the following three requirements:a) fast and accurate detection system
b) rapid alert system to inform affected parties
c) effective response mechanism
The solution being marketed by Digital Impact comprises three tools that meet
each of the above requirements.
Fraud Detection-Alert-Response system (Fraud-DAR) comprises of
Phishing Site Detection, Email Phish Detection, ISP Alert and Response,
Customer Alert System and Real Time Reporting elements. The system
allows the user to detect a possible attack early and inform potential victims.
Personal Authentication System is made up of the Recipient-Selected
Authenticity Phrase and Sender-Selected Authenticity Phrase elements. This
system allows the user to distinguish legitimate mail from fraudulent email
because each email can include a unique, personal authentication item, for
example an image, code, phrase, rewards balance, etc..
Email Verification Database comprises the Customer Service Email
Database which allows the user to query the database using the cusomer’s
email address and email code identifier and quickly determine whether an
email was sent by the company or is fraudulent.
19
Scanning and alerting software to detect and defend
against phishing attacks
All scanning and alerting mechanism look for patterns of previous attacks and
attempt to stop any future attacks. This section explores intrusion detection
systems for email filtering and web browser support
Email auditing is the process of checking the email after the actual
transmission. Email filtering is the process of intercepting and checking email
during transmission.
Email auditing can be achieved by using intrusion detection systems. An
intrusion detection system automates the monitoring and analysis process. As
this is a reactive, process, the monitoring and analysis will occur after the
email is sent.
An intrusion detection system (IDS) is a device that monitors activity to identify
malicious or suspicious events.
Intrusion detection systems are either signature or heuristic based .Signature
based systems perform simple patter matching and report situations that
match a pattern corresponding to a known attack type. Heuristic systems build
a model of acceptable behaviour and flag exceptions to that model.
Intrusion detection devices can either be host or network based. Host based
systems operator on a host to detect malicious activity on that host. Network
based systems operate on network data flows
Intrusion detection software builds patterns of normal system usage triggering
an alarm any time the usage seems abnormal.
An active intrusion detection system may be used to get the signatures of
known phishers and then the IP can be blocked. IP’s can also be obtained by
the use of a honeypot.
20
“Honeypots simulate one or more network services, hoping that an attacker
will attempt an intrusion. A honeypot is configured to interact with potential
hackers in such a way as to capture the details of their attacks. A properly
configured honeypot monitors traffic passively, doesn’t advertise its presence,
and provides a preserved prosecution rail for law enforcement agencies.“
(Lininger and Dean 2005).
Email auditing can also be achieved by scanning audit logs of email activity.
Logging is a reactive process which can occur at the application, host or
network level, meaning that an email trail can be picked up at any point in the
communication chain.
Most email systems use access control and /or passwords to authenticate the
user and determine what activity the user can perform. The use of access
control and passwords can be audited at the application, host and network
level hence the users activity can be monitored.
Email filtering can be achieved by the use of firewalls where specific data is
either blocked or allowed to go through. The firewall can either be
implemented at the host or network level. At the host level filtering occur for
one server whereas at the network level filtering occurs for all servers behind
that firewall. Personal firewalls can also be used for filtering on a per user
basis.
Phishing, also known as web spoofing is considered by Chou et. al. as a
special case of intrusion detection and they propose a browser plugin (called
SpoofGuard) to help combat web spoofing.
SpoofGuard is an Internet Explorer (IE) plugin that accessed the IE history file
and three other files stored in the user profile directory:1. file of host names of email sites
2. file of hashed password history
3. file of hashed image history
21
Below s a diagrammatic representation of the SpoofGuard architecture from
Chou et. al., followed by a summary of the characteristics of the SpoofGuard
architecture:-
Figure 12. SpoofGuard architecture
-
SpoofGuard uses domain name, url, link and image checks to evaluate
whether a page is being used in a spoof attack
-
SpoofGuard uses history logs to make intelligent decisions about user
behaviour
-
SpoofGuard intercepts and evaluates user posts
-
SpoofGuard calculates and evaluates the spoof index of a page
-
SpoofGuard compares post data to previously entered passwords from
different domains
When the spoof index is above a user-specified threshold a pop up window is
displayed with an additional warning. The user then has the choice to stop
what they are doing for continue sending data to the third party site. If the user
does not like pop ups a less intrusive tool bar is also available to give the user
information about the page.
22
Figure 13. SpoofGuard toolbar
The Options button allows the user to configure the tool. The traffic light (red,
yellow, green) gives an indication about the current page. If the user clicks the
traffic light then additional information about the page is displayed.
Chou et. al. evaluated SpoofGuard against similar characteristics as intrusion
detection systems and found the results very favourable.
Email verification tools
According to the Anti-Phishing Working Group (APWG) as quoted in Lininger
and Dean (2005), a two step email authentication standard could stop 85% of
phihsing attacks in their current form.
Lininger and Dean (2005) list the following four as the main contenders for
authentication:ƒ
SPF (Sender Policy Framework)
Checks the ‘envelope sender’ of an email message-the domain name of
the initiating SMTP server
ƒ
Sender-ID
Checks after the message data is transmitted and examines several
sender-related fields in the header of an email message to identify the
purported responsible address
ƒ
DomainKeys
Checks a header containing a digital signature of the message. It verifies
the domain of each email sender as well as the integrity of the message.
ƒ
IIM (Cisco Identified Internet Mail)
Adds two headers to the RFC 2822 message format to confirm the
authenticity of the senders address.
Sender Policy Framework (SPF) was formerly known as Sender Permitted
Form and is an extension of the Simple Mail Transfer Protocol (SMTP).
23
According to Lininger and Dean (2005) “When a user sends you mail, an
email servers connects to your email server. When the message comes in,
your mail servers can, based on SPF published addresses of its email
servers, tell if the server on the other end of the connection actually belongs to
the sender.”
Sender-ID, proposed by Microsoft, is the culmination of two previous
protocols, namely Caller ID and Sender Policy Framework (SPF). The aim
of the Sender-ID protocol is to only allow authenticated messages to reach the
receiver. According the Germain (2005) Sender-ID attempts to achieve its aim
in the following manner. The sender sends an email message to the receivers
inbound mail server. The receiver’s server checks for a record of the sending
domain published in the DNS record. The inbound email server determines if
the sending email server’s IP address matches the IP address published in
the DNS record.
DomainKeys is used to verify the domain of the email sender and the integrity
of the message. DomainKeys uses public key encryption technology to
achieve the verification. The process of how DomainKeys are used can be
described in two parts, Sending DomainKey email and Receiving DomainKey
email according to Lininger and Dean (2005) .
Sending DomainKey email involves performing a secure hash of the email
contents using the SHA-1 algorithm; encrypting the result using a private key
with RSA algorithm; and encoding the encrypted data using Base 64.
Receiving DomainKey email involves the server using the name of the domain
from which the email originated to perform a DNS lookup to get that domain’s
public key. The receiver decrypts the hash value in the header and
recalculates the hash of the body. If the two values match then the receiver
can be confident of the origin of the email.
When comparing Sender-ID and DomainKeys Gillis as quoted in (Germain
2005) commented that “Sender-ID is very lightweight but is already very
widespread. Domain Keys is definitely stronger because of the encrypted
signature.”
24
Cisco Identified Internet Mail (IIM) was designed by Cisco System to help
identify fraudulent email. IIM is a signature based email authentication
standard that adds two headers to the message format: IIM-Signature and
IIM-Verification. Lininger and Dean (2005) state that “To establish the
authenticity of an email message, IIM verifies that the message sender is
authorized to send messages using a given email address and that the
original message was not altered in any consequential manner.”
Website verification Solutions
Fraudsters prey upon the trusted brand relationship well known companies
have with their customers; Website Verification Solutions try to answer the
question of how to distinguish authentic web sites from copycats.
Client Side Solutions
Ollman (2004) describes the following three hidden attacks that are used to
manipulate the display on the victims web browser:- hidden frames; overriding
page content and graphical substitution. One small step towards countering
these threats is to be aware of the browser specific visual clues of graphical
substitution that Ollman (2004) lists, viz. “…the URL presented within the
browsers URL field, the secure padlock representing an HTTPS encrypted
connection, and the zone of the page source.”
The fraudulent website has to either be a full copy of the legitimate site or
make reference to the real site for part of the graphics/content.
Developers can make the fraudster’s task much more difficult by employing
Image Cycling as suggested by Ollman (2004). Image Cycling is the
technique of uniquely naming and recycling images periodically. Any
fraudulent site that references an image by the old name could have content
supplied that displays a warning to the user. A variation of this technique is to
extend the name to include the user session.
25
Another graphical security technique is to use an image hash rather than the
actual image on the legitimate site, Chou et. al. (2005)
Ollman (2004) also describes the use of “…agent-based bots to monitor
URL’s and web content from remote sites, actively searching for all instances
of an organisations logo, trademark or unique web content.” Obviously once
the fraudulent website content is detected, the owner can take remedial/legal
action.
The Comodo Website describes Content Verification Certificates (CVC’s) as
an effective means of verifying website content. The website content is first
checked to ensure that it’s verifiable (see requirements below) and then an
X509 compliant certificate is issued. There is a high degree of confidence in
the CVC’s as they are built on the public key infrastructure. The following are
requirements for web site content to be verifiable according to the Comodo
website:ƒ
Suitably complex such that it cannot easily be spoofed (No cut and
paste possibilities)
ƒ
Directly linked (bound) to the web page (URL and or IP) upon which it
is to be displayed
ƒ
Given a validity period related to its usage.
Chou et. al. (2005) presented three groups of client side tests that can be
used to distinguish spoof pages from legitimate pages, i.e. stateless, stateful
and methods that evaluate outgoing html post data:Stateless methods are be used to evaluate the current page to determine if it
is suspicious or not. Examples of stateful methods would be: - The URL of the
page is checked for correct form and any links on that page are also checked.
Any page that requests a password can be checked for https and whether the
site is using a valid certificate.
Stateful methods are used to evaluate a page based on previous user activity.
The domain of a page is checked for a match against previously visited pages
26
or a close match in the domain name. Chou et. al. (2005) achieves this check
by calculating the Hamming (edit) distance. The referrer page can also be
used as an indicator. As most phishing attacks are initiated by email, if the
referrer is an email site then this raises the suspicion level.
Methods that evaluate outgoing html post data includes hashing and storing
sensitive data in a database so that any outgoing post data is checked against
the password hash to detect password leakage.
Server Side Solutions
Chou et. al. (2005) describes the following server side methods that can be
added to the arsenal for website verification:Marking form fields with confidentiality tags will allow security tools to track the
correct information thereby reducing false alarm rate. The confidentiality tag
can also be used by browsers to flush cookies if the user had visited a site
that required confidential data input
Image tagging is implemented by adding a tag to the page to mark the image
as not to be used on pages other than pages belonging to the domain. If the
tag is found on pages outside the domain then the page is flagged as
suspicious.
The use of password hashing and site specific salt will produce distinct
passwords for distinct sites. The salt has to be a unique value and Chou et. al.
(2005) recommends using the site domain name as the salt. Since users tend
to use the same password across sites, once a phisher has cracked one site
they try the same password on other sites so password hashing and the site
specific salt is an effective countermeasure.
Another techniques that Chou et. al. (2005) describe but are not in favour of
are collaborative methods that rely on the user informing a central server of a
spoofed site, the central server then alerts all plug-ins to block this page.
HIPs
The implementation of the Human Interactive Proofs (HIPs) criteria proposed
by Rachna and Tygar (2005) is called Dynamic Security Skins (DSS) and
provides a solution that allows the “remote server to prove it’s identity in a way
27
that is easy for a human user to verify and hard for an attacker to spoof.”
(Rachna and Tygar, 2005:p136)
According to Rachna and Tygar (2005) “… (HIPs) allow a computer to
distinguish a specific class of humans over a network. “. They extend the
definition of HIPs to allow a user to issue a challenge to the computer. The
characteristics of the challenge are:ƒ
“be easy for a particular class of computer to pass
ƒ
be hard for other computers to pass, even after observing a number of
successful authentications
ƒ
produce results that are easy for a human to verify
ƒ
use a protocol this is publicly available
ƒ
not require the user to have specialized tools”
Guarantee of legitimacy of messages and their source
To guarantee the legitimacy of messages once must look at the question of
determining whether the website content is from the provider who the site
purports is the provider. Fraudster use many tactics to spoof a site, including
- official looking and sounding email
- copies of legitimate corporate emails with minor url changes
- use of fake “Mail From:” addresses and open mail relays for disguising the
source fo the email
The popularity of Instant Relay Chat (IRC) and Instant Messaging (IM) has led
to these applications being enriched with dynamic embedded content, for
example URL’s and graphics. Many of the phishing tricks used on web sites
can now we used on IRC and IM applications as well.
Validating Official Communication
Official communication sent be an organisation can be constructed in such a
manner as to give the recipient visual clues to its authenticity or lack thereof.
Some methods of helping the recipient judge the authenticity of a message,
identified by Ollman (2004), are email personalisation; visual/audio
28
personalisation of messages; previous message referral and digital
signatures.
Email personalisation
The organisation personalises any email with information that is only shared
between the organisation and its customer, hence generic email will
immediately cause the customer to be suspicious. However the organisation
has to vigilant about who can access to the shared customer information. It’s
possible to use visual and audio data for the purposes of personalisation.
Previous Message Referral
The organisation can, in current correspondence reference the last
correspondence with its customer. Ollman (2004) identifies the following
suggestions:ƒ
Clearly referencing the subject and date of the previous email
ƒ
Sequentially numbering the emails
Digitally Signed Email
By digitally signing the message the recipient can be confident of the
originator, i.e. the organisation/person that signed the email, leading to a
better trust relationship. However the onus is on the recipient to verify the
signature thereby ascertaining the true source of the message. Ollman (2004)
issues a word of caution though, there is nothing stopping the phisher from
creating a public/private key pair and digitally signing his mail.
Mail Server Authentication
There are 2 ways to ensure you only communicate with message senders that
are trusted according to Ollman (2004). Either you verify the IP or range by
executing a reverse resolution of Domain information or you use secure
SMTP. The first method will cause emails to be dropped if the sending domain
cannot be verified while the second method will fail at the establishment of a
secure connection. The purpose
29
One of the best practices recommended by ASTA (2004) to ensure that the
source of a message is known is to reconfigure any mail servers that are
configured as open relays to be secure relays. Open relays allow messages to
pass through the mail server obscuring the sender from the receiver.
Another recommendation from the Anti-Spam Technical Alliance (2004) is for
ISP’s are to ensure anyone wanting to send email must have a valid account
on the system. The ISP’s can also impose rate limitations on outbound mail
traffic to curb spam that is being used as a device to spread viruses/worms.
Anti-Spam Technical Alliance (2004) suggests “The limits should be based on
To/Cc/Bcc recipient counts per unit of time from end user account or server IP
address.”
Domain spoofing obscures the origin and true sender of email by forging the
sender address and domain name. It is widely acknowledged that the IP
address is the only trustworthy element of the email header. Anti-Spam
Technical Alliance (2004) suggests one approach to counter domain spoofing
is to validate the domain information in the header against the IP of the
domain. The IP is them matched against a publicly known list of IP that are
allowed to send email on behalf of that domain, obviously any deviation will
create suspicion of a security risk.
Anti-Spam Technical Alliance (2004) also suggested Content Signing (CS)
technology will be valuable in verifying the sender’s identity and the message
contents. CS makes use of public key/private key pairs to digitally sign
messages and achieve the authentication.
Two factor authentication for customer access to
financial services
The case for two factor authentication is strongly stated by Ollman (2004) as
“… to clearly determine consumer identities so online businesses can avoid
the costs of being defrauded and dramatically reduce the overhead costs of
today’s manual methods of fraud and theft prevention.”
30
Two-factor authentication technology refers to the dual authentication
mechanism of something you know (a password) and something you possess
(a token). The something you possess could be a physical device, (like a
smartcard or key-fob) or a single-use or time-dependant password.
The purpose of two factor authentication, according to Ollman (2004) is to
“…create strong (one-time) passwords that cannot be repeatedly used to gain
entry to an application.”
Figure 14. Strong token-based authentication (Ollman 2004, p.34)
Hardware tokens devices currently available are of the Challenge-Response
method or the SecureID devices from RSA security described below.
Kumar (year unknown) describes client certificates as being able to provide
strong authentication in a web application. A Smart Card implementation of
client certificates provides “a secure and mobile platform for authentication”,
Kumar (year unknown).
Example: RSA SecureID Consumer Protection Solution
RSA Security Inc white paper (2004) describes the RSA SecureID Consumer
Protection Solution implementation of two factor authentications in the
following manner:- A hardware device, the RSA SecurID authenticator is issued to each person
that needs secure access
- The RSA SecurID authenticator generates a new, unpredictable code every
60 seconds.
31
- Two factor authentication is achieved by combining the authenticator with a
PIN
- Persons needing secure access then combine something they know, i.e. the
PIN with something they possess, i.e. the constantly changing code on the
authenticator
The benefits of two factor authentication in the context of the RSA SecureID
Consumer Protection Solution are as follows:- The one time password changes every 60 seconds so a phisher cannot steal
and use old codes
- There is no requirement to install additional software
- The authenticator is mobile hence secure access can be achieved from any
device connected to the internet
- Users don’t have to remember multiple passwords
- Businesses can better protect themselves against fraud hence reestablishing trust in e-commerce
Schneier (2005) argues that two factor authentication is no longer adequate
protection when transacting over the internet as “the real threat is fraud due to
impersonation…Two factor authentication will force criminals to modify their
attacks, that’s all”. Criminals have modified their approach by using Man-inthe-middle and Trojan attacks to render two factor authentications useless.
Man in the middle attacks
The attacker places an intermediary between the victim’s browser and the real
web server to take control of the victim’s resources and confidential data and
proxies all communication. Ollman (2004) lists the following 4 attack vectors
for Man-in-the-middle attacks:- Tranparent Proxies; DNS Cache Poisoning;
URL Obfuscation and Browser Proxy Configuration
32
Figure 15. Man-in-the-middle attack structure (Ollman 2004, p9)
The man-in-the-middle attacks are successful because the attacker is
proxying all communication between the users and secure resource so the
attacker can easily insert their own transactions along with the user initiated
transactions.
Trojan attacks are successful because the attacker uses the Trojan to piggy
backs on the user’s legitimate session with the secure resource.
The use of strong two factor authentication opens up opportunities for creating
communities of user and sites that trust each other or as RSA Security Inc
white paper (2004) describes them “federated consumer identity protection”.
The benefits of creating such trusted domains are:ƒ
shared cost of implementing security across multiple systems
ƒ
ease of and decreased time to market for a new site as they can now
leverage of a proved authentication system
ƒ
consumers have one way of authenticating themselves across multiple
systems
Two phase login
Ollman (2004) describes a two phase login process that can make the
authentication process more secure. Phase one involves the input of data that
may be common knowledge, e.g. the account number and login name. Once
the user gets past phase one, they are required to input 2 or more unique
pieces of authentication information. Phase two can also present a
personalised graphic that potentially acts as a watermark on that page.
33
Identity Theft
Van Oorschot and Stubbebine (2004) define identity theft as “… the
unauthorised use and exploitation of another individual’s identity-corroborating
information (e.g. name, home address, phone number, social security
number, bank account numbers, etc.)” The purpose of this type of crimes is to
obtain enough information to either apply for new official documentation in the
victim’s name or use the existence of the identity of the victim to conduct
illegal activities.
The Identity Theft Resource Centre (2005) cites Gartner Research and Harris
Interactive who estimate approximately 7 million people falling victim to
identity theft in the prior 12 months.
They also cite a GAO study on the costs of identity theft in America
“A GAO study on identity theft (GAO-02-363, issued March 2002) discussed costs to federal
agencies -- The executive office for U.S. Attorneys estimated cost of prosecuting a whitecollar crime case was $11,443. The Secret Service estimates the average cost per financial
crime investigation is $15,000.The FBI estimates the average cost per financial crime
investigation is $20,000”
These 2 statistics reinforces the severity of the problem and the seriousness
with which counter measures need to be sought.
Identity theft is becoming increasingly popular because it is a relatively easy
crime to commit and the rewards (monetary) are substantial.
Van Oorschot and Stubbebine (2004) believe that the reason personal and
financial information is so easily stolen is it’s very easy for the fraudster to
duplicate personal information but very difficult for the victim to know that a
duplicate has been made. When the fraudster applies for duplicate
documentation using stolen personal information there is no mechanism for
the legitimate owner to know that this process is taking place.
The Special Interest Group (SIG) on Identity Theft in the Liberty Alliance
Whitepaper: Identity Theft Primer (2005) have categorised identity theft into
the following three groups:-
34
ƒ
‘True’ name identity theft - the fraudsters use stolen personal
information to obtain new accounts and services
ƒ
Account takeover – the fraudsters use the stolen information to gain
access to the victims legitimate accounts and services
ƒ
Criminal identity theft – the fraudsters use the stolen personal
information to evade having criminal record in their own name
For Identity theft to be viable the information must first be accessed and then
used for nefarious purposes. The SIG on Identity Theft believe that identity
theft is not an ad hoc crime of convenience, and can be effectively used in a
planned attack using the steps in the six phases described below
Figure 16. . Identity Theft Lifecycle (Liberty Alliance Project 2005, p.7)
Phishing, key loggers and screen grabbers have proven to be effective means
for the fraudster to obtain this “identity-corroborating information”. Users
readily disclose this information on a spoofed site as there is inherent trust in
the requesting authority, i.e. the bank whose logo/banners are prominently
displayed.
35
Van Oorschot and Stubbebin (2004) believe key logging now rivals phishing
as the tool of choice to illegally obtain personal information. Lemos (2004) as
cited in Van Oorschot and Stubbebine (2004) describe the example of the
Bankhook.A Trojan that recorded sensitive information prior to SSL encryption
and mailed the data to a remote computer.
The Liberty Alliance Project (2005) advocates the three pronged approach of
technical, operational and policy countermeasures to identity theft. Two
technical approaches are described below, the use of activity monitoring and
the theoretical architecture proposed by Van Oorschot and Stubbebine
(2004). The operational approach necessitates the agreement to and
widespread implementation of contracts and best practices. The policy
countermeasure requires the legal and regulatory compliance of institutions
that store and process personal information.
Many systems and companies attempt to recognize identity theft by
monitoring activity/transactions made by an individual. The occurrence of
transactions that fall outside the normal pattern of behaviour is then flagged
as suspicious and warrants closer inspection. This approach is effective if the
normal pattern of transacting is well defined and we are dealing with a
transaction based system like credit card activity.
The solution proposed by Van Oorschot and Stubbebine (2004) is an
architecture that combines “… a physical location cross-check, a method for
assuring uniqueness of location claims, and a centralized verification
process.” The following is an overview of the proposed system:ƒ
Every user has some device that is used to securely determine their
location; Van Oorschot and Stubbebine (2004) refer to this as the
“heartbeat locator”. Examples of heartbeat locators would be cell phone
and wireless personal digital assistant (PDA).
ƒ
To address the concern of multiple “identities” of the same person
being created every identity verification goes through a centralized
system that monitors for anomalies. Van Oorschot and Stubbebine
(2004) refer to this property as “entity uniqueness”
36
ƒ
At transaction time, the location of the transaction is matched to the
location of the person (heartbeat locator) to create a real time, online
verification system
Conclusion
This report gave a brief introduction to the phishing problem and background
information on how an attack is executed. A selection of countermeasures
was then discussed: - the EarthLink and eBay browser plugins as viable client
side defences; scanning and alerting software was also looked at as an
alternative to user intervention and email verification tools to counter spam.
The section on website verification solutions looked at the current attack
vectors that are being used to create fake websites and their
countermeasures. There was also a discussion on how to guarantee the
legitimacy of messages and their source.
Two factor authentication was described as well as a commercially available
implementation of this idea. An argument against the effectiveness of two
factor authentication was also provided. Lastly there was brief discussion of
identity theft and some proposals on how to handle this growing problem.
Technology alone will not stop what has become a lucrative activity. Greater
awareness is needed and users must be educated about the risks associated
with divulging personal information online. Businesses must also protect the
user’s information more securely for example by ensuring that information is
never transmitted in clear text.
There is also a need for greater co-operation and information sharing between
financial institutions, government agencies and technology providers.
And if we ever want to provide a deterrent to this activity then we need better
trained law enforcement agents to find and prosecute the perpetrators of
these crimes.
37
In conclusion one can see that there are many attempts to propose and
implement solutions to the phishing phenomenon. However the problem
domain is constantly changing, as new solutions are found, the attack vectors
are modified to become more sophisticated.
There is a long road ahead before the tide of phishing attacks is stemmed, if
indeed that is possible.
38
References
Anti-Phishing Working Group 2005, The Anit-Phishing Working Group, viewed
31 October 2005, ,<http://www.antiphishing.org/>
Anti-Spam Technical Alliance (ASTA) 2004, Anti-Spam Technical Alliance
Technology and Policy Proposal, viewed 10 February 2006, <
http://docs.yahoo.com/docs/pr/pdf/asta_soi.pdf>
Chou, N, Ledesma, R, Teraguchi, Y, Mitchell JC ‘Client-side defense against
web-based identity theft’, Computer Science Department, Stanford Univeristy,
viewed 07 November 2005, <
http://www.isoc.org/isoc/conferences/ndss/04/proceedings/Papers/Chou.pdf>
DigitalImpact.com 2005, Digital Impact Anti-Phishing Solutions, viewed 29
October 2005, <
http://www.digitalimpact.com/pdfs/AnitPhishing_Solutions_Data_Sheet.pdf >
Drake, Christine, E., Oliver, J., Koontz, Eugene, J. “Anatomy of a Phishing
Email”. 2004
Facts and Statistics 2005, Identity Theft Resource Center, viewed 10
February 2006, < http://www.idtheftcenter.org/facts.shtml>
FDIC – Identity Theft Symposium: Fighting Back against Phishing and
Account Hijacking. “What More Can Be Done to Fight Phishing: The Indistry
Perspective” . 11 February 2005
Germain, Jack M 2005, “Adoption of Alternate E-Mail Security Technologies
Pushed”, TechNewsWorldI, viewed 29 October 2005, <
http://www.technewsworld.com/story/44156.html >
Hall, Kirk. “Vulnerability of First-Generation Digital Certificates and Potential
for Phishing Attacks and Consumer Fraud (GeoTrust Whitepaper)”. April
2005.
Hess, A, Seamons, KE 2003, ‘An Access Control Model for Dynamic ClientSide Content’, viewed 10 February 2006, <
http://isrl.cs.byu.edu/pubs/sacmat2003.pdf>
Jakobsson, Markus. “Modeling and Preventing Phishing Attacks”. Phishing
Panel in Financial Cryptography '05. 2005.
39
Kerner, S.M. “The Cost of Phishing Hits $1.2 Billion”, 6 May 2004
www.internetnews.com/ec-news/article.php/3350891
Kumar, A (year unknown), ‘Phishing – A new age weapon’, OWASP The
Open Web Application Security Project, viewed 10 February 2006, <
http://www.paladion.net/papers/Phishing_a_new_age_weapon.pdf >
Liberty Alliance Whitepaper: Identity Theft Primer, 2005, viewed 10 February
2006, <https://www.projectliberty.org/resources/id_Theft_Primer_Final.pdf>
Phishjing activity trends report February 2005. Anti-Phishing Working Group
(APWG), viewed 1 April 2005, <
http://antiphishing.org/APWG_Phishing_Activity_Report_February_2005.pdf>
Paypal- 'Unauthorized Account Access 10 May 2005 Anti-Phishing Working
Group (APWG)., <http://www.antiphishing.org/phishing_archive/05-1005_Paypal/05-10-05_Paypal.html>
Madsen, P 2005, ‘Liberty ID-WSF People Service – federated social identity
’, Liberty Alliance Project White Paper, viewed 10 February 2006, <
https://www.projectliberty.org/resources/whitepapers/Liberty_Federated_Socia
l_Identity.pdf>
Morrison, Scott., Nuttall Chris. “Gone phishing Fraud based on bogus e-mail
is causing concern among banks and online retailers” (FT Report – FT-IT
Review). 23 June 2004
Ollman, G 2004, Understanding & Preventing Phishing Attacks, The Phishing
Guide, NGSSoftware Insight Security Research, viewed 10 February 2006, <
http://www.ngssoftware.com/papers/NISR-WP-Phishing.pdf>
Pfleeger, C. & Pfleeger, S.: Security in Computing. Upper Saddle River, NJ:
Prentice-Hall (2003) 3 ed.
Phishjing Activity Trends Report Novenmber 2006. Anti-Phishing Working
Group (APWG), viewed 10 February 2006, <
http://antiphishing.org/reports/apwg_report_Nov2005_FINAL.pdf>
Protecting Against Phishing by Implementing Strong Two Factor
Authentication 2004, RSA Security Inc white paper, viewed 10 February 2006,
< http://www.antiphishing.org/sponsors_technical_papers/PHISH_WP_0904>
Rachna D, Tygar J.D 2005, ‘Phish and HIPs: Human Interactive Proofs to
Detect Phishing Attacks’, viwed 10 February 2006, <
http://www.cs.berkeley.edu/~tygar/papers/Phish_and_HIPs.pdf>
40
Rudd, Barney. “An analysis of Phishing and possible mitigation strategies”. 27
June 2004
Schneier, B 2005, Two-Factor Authentication: Too Little, Too Late, viewed 10
February 2006, < http://www.schneier.com/essay-083.html>
Shor, Susan B 2005, “Cisco, Yahoo Propose Anti-Spam Technology
Standard”, TechNewsWorld, viewed 29 October 2005, <
http://www.technewsworld.com/story/44574.html>
Tumbleweed Communications. “Using Digital Signatures to Secure Email and
Stop Phishing Attacks (A Tumbleweed Whitepaper)”. 2004
Tynan, Dan 2004, “EarthLink Readies Anti-Phishing Tool ScamBlocker is
designed to track, block access to known swindle sites”, PC World, viewed 29
October 2005, <http://yahoo.pcworld.com/yahoo/article/0,aid,115652,00.asp>
Van Oorschot, P.C, Stubbebine, S 2004, Countering Identity Theft through
Digital Uniqueness, Location Cross-Checking and Funneling, viewed 10
February 2006, < http://fc05.ifca.ai/p03.pdf>
What are CVCs? – Content Verification Certificates 2005, Content
Verification, viewed 10 February 2006, <
http://www.contentverification.com/cvc/index.html>
41